Decision intelligence

One recommendation. Its full evidence. Or an honest refusal.

Aegis reads your operating telemetry, finds the controllable driver behind the outcome you care about, and states what to act on. When the evidence cannot support a judgment, it says so instead of inventing one.

Deterministic
Same data, same judgment, every run
Explainable
Proof graph and validator rules on every decision
Defensive
Refuses rather than guesses on weak evidence

What Aegis does

It replaces the analysis meeting, not the analyst.

Aegis does the part executives cannot delegate reliably: separating the metric that moved from the metric that caused it.

Structure

Reads any operating dataset

Metrics are profiled by shape, unit, trend and topology. Aegis infers operational roles — driver, outcome, constraint — without a metric dictionary.

Causality

Ranks controllable drivers

A proof graph ranks candidate causes by controllability, structural weight, lead evidence and trend strength, then targets the strongest upstream cause.

Restraint

Knows when to stop

Coverage gates, oscillation gates and premise-conflict evaluation block a recommendation whenever the evidence does not carry it.

How it works

Eight deterministic stages, all inspectable.

Nothing is hidden behind a model. Each stage produces an artefact you can open.

  1. 01

    Upload dataset

  2. 02

    Telemetry profiling

  3. 03

    Coverage analysis

  4. 04

    Proof graph

  5. 05

    Driver selection

  6. 06

    Business judgment

  7. 07

    Validation

  8. 08

    Executive report

Why trust Aegis

A judgment you can defend in a board meeting.

Certified against adversarial scenarios

A fixed benchmark of cross-domain scenarios with hidden oracles scores the engine on premise handling, root-cause precision and refusal behaviour.

No training on customer telemetry

Your data is never used to improve the engine. There is no model to improve — the reasoning is rules and structure, and it is versioned.

Every decision is recorded

Recommendation, evidence, confidence, whether you followed it and what happened. Accountability that survives staff turnover.

Explainable by construction

Aegis cannot produce a recommendation without a traceable path from telemetry to driver to judgment.

Industries supported

Reasoning that generalises across operating models.

SaaS

Retention, expansion, onboarding and support telemetry.

Manufacturing

Yield, defect rate, calibration intervals, downtime.

Healthcare

Capacity, staffing, wait time, readmission.

Retail

Footfall, conversion, stockouts, replenishment.

Logistics

On-time delivery, dwell time, fleet maintenance.

Professional services

Utilisation, delivery quality, margin.

Executive decision workflow

From export to accountable decision.

  1. 01

    Upload a snapshot

    Any CSV export with periods across columns. No schema, no mapping, no integration project.

  2. 02

    Read one judgment

    A single recommendation aimed at the controllable upstream driver — not the symptom.

  3. 03

    Open the evidence

    Every metric, edge and validator rule that produced the judgment is on the page.

  4. 04

    Record the outcome

    What you did, what happened. The decision ledger becomes institutional memory.

Testimonials

Pilot feedback, published once it exists.

Aegis is in executive validation. Rather than fabricate quotes, this space stays empty until pilot executives approve their own words.

Manufacturing pilot

Reserved for a named executive quote after the validation period closes.

Healthcare pilot

Reserved for a named executive quote after the validation period closes.

SaaS pilot

Reserved for a named executive quote after the validation period closes.

FAQ

Direct answers.

Is this a chatbot?

No. Aegis has no conversational surface and no generated prose. The engine is deterministic: the same dataset always produces the same judgment, evidence chain and confidence.

What happens when the data is weak?

Aegis refuses. Insufficient coverage, oscillating telemetry with no trend, or a premise contradicted by the evidence all resolve to HOLD_JUDGMENT rather than a confident guess.

Does it work outside software businesses?

Yes. Roles are inferred from the structure of the telemetry — shape, units, lead and lag — not from a dictionary of metric names. Manufacturing, healthcare, retail and logistics datasets are certified in the benchmark.

Where does my data go?

Analyses run in your browser. No dataset is uploaded to a server and no customer telemetry is used to train anything.

How do I know the reasoning is sound?

Every judgment exposes its proof graph, driver ranking and the validator rules that fired. The engine is scored against a fixed benchmark of adversarial scenarios with hidden oracles.

See Aegis judge your own numbers.

Bring one CSV export. In a 30-minute session we run it live and walk the full evidence chain behind whatever Aegis concludes — including a refusal.